Optimization of Functional MRI Techniques for amos3.aapm.org/abstracts/pdf/68-19722-230349-90274.pdf

  • View

  • Download

Embed Size (px)

Text of Optimization of Functional MRI Techniques for amos3.aapm.org/abstracts/pdf/68-19722-230349-90274.pdf

  • 8/2/2012


    Optimization of Functional MRI Techniques

    for Tumor and Normal Tissue Response to RT

    Yue Cao, Ph.D. Departments of Radiation Oncology,

    Radiology and Biomedical Engineering University of Michigan1

    Acknowledgments n Radiation Oncology n James Balter, Ph.D. n Edgar Ben-Josef, MD n Avraham Eisbruch, MD n Mary Feng, M.D. n Felix Feng, MD n Theodore S. Lawrence, MD, Ph.D n Charlie Pan, MD n Randall Ten Haken, Ph.D. n Christina I. Tsien, MD n Clinical Coordinators

    Alan AntonuK, BS Chris Chapman, B.S. Reza Farjam, M.S. Mohammad Nazemzadeh, Ph.D. Hesheng Wang, Ph.D. Peng Wang, Ph.D.

    n Radiology n Thomas L. Chenevert, Ph.D. n Hero Hussain, M.D. n Diana Gomez-Hassan, M.D., Ph.D. n Suresh Mukherji, MD n Pia Maly Sundgren, MD, Ph.D. n Radiology staffs

    n STATISTICS n Tim Johnson, Ph.D. n Dan Normolle, Ph.D. n Matt Schipper, Ph.D.

    n NIH grants – 3 PO1 CA059827 (Ten Haken) – RO1 NS064973 (Cao) – RO1 CA132834 (Cao)


    Biological Target Volume, Tx Predictor, Dose Panting and Sculpting


    Ling, IJROPB, 2000

    Dose painting

    & sculpting

    Tx Assessment

    • A tumor target volume could be defined and segmented as multiple biological target subvolumes, which are defined based upon multiple functional imaging stuidies, and each of which could be a prognostic or predictive indicator for radiation response.

    • Dose sculpting and painting of multiple biological target subvolumes could lead better outcome.

  • 8/2/2012


    Imaging Biomarker for Therapy Assessment

    Ø Longitudinal study • Repeated measures • Reliability, reproducibility, robustness

    Ø Sensitive indicator (biomarker) • Sensitive to therapy-induced changes

    Ø Prognostic or Predictive biomarker • predict Tx response and outcome • surrogate endpoint


    Physiological MRI




    CBVHepatic Perfusion



    How to Establish a Imaging Biomarker for Therapy Assessment

    Ø Reproducibility • Separation of a true change from variation

    Ø Sensitivity and specificity • end points, problem-specific

    Ø Utility • therapy selection • adaptive therapy for intensification or toxicity



  • 8/2/2012



    Ø NCI RIDER project • A set of articles describe importance, procedures

    and statistical analysis for test and re-test (Translational Oncology, 2, 2009)

    • Test and retest data (NBIA database)

    Ø What level of change in parameter should be observed to be confident that there has been a true change in the parameter in an individual patient?


    How to determine repeatability from Test and Retest Data

    Ø Within-subject SDw of a parameter Ø Repeatability Coefficient of a parameter

    • RC=2.77SDw Ø RC defines 95% confidence interval Ø A true change from a subject has to be

    • |upost-upre| >RC

    8 Barnharts et al, Translational Oncology, 2, 2009

    Repeatability of Segmenting Cinglum Fiber Track

    Ø Develop a method for segmenting cinglum, a fiber track connecting to hippocampus, for assessment of radiation-induced cognitive dysfunction

    Ø Evaluate reproducibility of segmented volumes and DTI indices

    dice coefficient: overlapped volumes RC: uncertainty of parameters


    Nazemzadeh, et a WED-G-217A-4 AAPM 2012

  • 8/2/2012


    Fractional Anisotropy Changes in Patients Received WBRT

    10Nazemzadeh, et al AAPM 2012

    95%CI Ture change


    -RC( W

    2- pr

    e R



    Posterior cinglum


    How to determine reproducibility of an image parameter






    18% 1. Use correlation coefficient 2. Use pair t test 3. Use repeatability coefficient 4. Use visual inspection 5. Ask a radiolgist

    Sensitivity and Specificity Ø Endpoints

    Prognostic or Predictive indicator for tumor response to therapy Clinical outcomes, local control, survival, PFS… Radiographic response Pathophysiology

    For normal tissue radiation toxicity Clinical outcome (symptom) Radiation dose Independently established measure 12

  • 8/2/2012


    Imaging-driven Response- Induced Subvolume of a Tumor

    ØHypothesis • Heterogeneous therapy response of a tumor could be

    primarily due to biological heterogeneity in the tumor • The most aggressive or resistant sub-volume in a tumor

    could predominantly determine therapy response or outcome of a tumor

    ØAim • Develop a methodology to delineate physiological

    imaging–driven subvolume of a tumor • Predictive for treatment response • Highly reproducible • Potential candidate to be a boost target 13

    How to identify sub-volumes

    Ø Generate a feature space of tumors defined by a joint histogram of physiological imaging parameters from “training” tumors

    Ø Classify the distribution of tumor features (joint histogram) using fuzzy c-means to determine globally-defined probability membership functions of the classes

    Ø Create the subvolumes (metric and map) of each individual tumor using the probability membership functions and physiological images of the tumor


    Joint histogram (feature space)

    Clustering analysis to

    determine PMF of classes

    Subvolumes of each tumor

    (metric and map)

    Test for Tx response


    Pr e-



    Brain metastases


    W ee

    k 2






    0.15T1 rCBV K trans

    Pr e-

    RT Responsive


    Farjam, et al AAMP 2012 WE-C-BRA-5

    probability membership functionCreate a single metric, a subvolume of a tumor with high CBV, Ktrans, or both, for assessment of response

  • 8/2/2012


    Does the tumor subvolume with high CBV predict response?

    Ø End point – Post-RT radiographic response • ∆GTVpost=GTV1mpost – GTVpreRT • Non-responsive: ∆GTVpost decreases

  • 8/2/2012


    Prediction of Local Failure


    Wang, et al Med Phys 2012

    Association with Pattern Failure


    Poorly-perfused Subvolume pre RT

    FDG 3 mon post RT

    Subvolumes of a Tumor Ø A physiological imaging defined response-induced

    subvolume of a tumor could be a candidate for intensified therapy target

    Ø Our approach can be applied to other physiological/metabolic imaging parameters

    Ø Our method does not depend upon the voxel level accuracy of registration of a pair of images acquired over a period of therapy

    Ø Our method produces the metrics robust to image noise and other random factors


  • 8/2/2012


    How to determine sensitivity and specificity of a parameter






    18% 1. Use a clinical end point & ROC analysis

    2. Recruit 500 patients

    3. Repeat the same study

    4. Ask a radiation oncologist

    5. Increase signal-to noise ratio of the image

    Normal Tissue Function

    Ø Radiation-induced tissue toxicity is a complex, dynamic and progressive process

    Ø Individual sensitivity to radiation limits the utility of dose-based NTCP models

    Ø Determination of specific risks versus benefits of treatment should be an integral part of clinical decision making for each patient


    How to assess normal tissue/organ function

    Ø Assess normal tissue/organ function • Dynamic changes in organ function from pre to during

    and to post RT • Spatially resolved volumetric distribution of function • Individual sensitivity to dose and pre-RT dysfunction

    Ø Early measures predict organ functional risks of radiation

    Ø Re-optimization of individual patients’ treatment plans and early intervention for tissue/organ protection


  • 8/2/2012


    MRI Perfusion Imaging to Assess Liver Function

    Ø Patients with high risks or poor liver function • HCC with or without cirrhosis • Large tumor • Previous treatments, TACE, surgery, RT, RFA

    Ø General paradigm • DCE MRI scans: pre, during (~50%), 1 month

    after RT


    Individual Perfusion Response


    Spatial perfusion changes are related to regional doses

    Cao, et al. IJROBP, in press Pre RT 1 mon Post RT

    Evaluation of Perfusion with Overall Liver Function


    r = 0.7 p

  • 8/2/2012


    Dose-Dependent Spatial Perfusion Changes



    Early Prediction of Post-RT Perfusion Change Ø Linear mixed effects model to account

    for individual random effects


    ijjijitjiidj edbdFaF +++++= βγα 0)log(

    Post-RT Perfusion for Subject i in region j received dose d

    Pre- or during-RT Perfusion for Subject i in region j

    Dose in region j

    Model 1: independent variable: pre-RT perfusion, total dose, predictor: total dose (p

  • 8/2/2012


    Localization and Detection of Prostate Cancer Using MRI

    Ø DCE, DW, T2W and MRS Ø DCE MRI is superior to dif